Washington Built the AI Infrastructure AUKUS Needs — Then Locked Allies Out

Key Takeaways

• US controls 74% of global AI compute capacity
• Genesis CRDAs only include American companies, exclude AUKUS partners
• Allies face billions‑cost domestic exascale builds or limited cloud
• Extending CRDAs would enable secure, joint AI model training
• Fragmented effort risks duplication and strategic lag behind China

Summary

The United States has built the Genesis supercomputing platform, giving American AI firms structured access through Cooperative Research and Development Agreements, but it excludes AUKUS partners. While Australia and the United Kingdom pour billions into quantum and autonomous defense technologies, they lack a comparable pathway to the exascale compute needed for Pillar II. The resulting gap forces allies to consider costly domestic builds, ad‑hoc cloud solutions, or fragmented development, undermining the alliance’s integrated innovation goals. Extending the Genesis CRDA framework to AUKUS could close the compute divide and preserve strategic cohesion.

Pulse Analysis

Genesis represents a rare convergence of U.S. policy and massive exascale compute, positioning the Department of Energy as the de‑facto hub for defense‑grade AI training. By channeling access through Cooperative Research and Development Agreements, the administration ensures tight IP control, export‑compliance, and security oversight for domestic firms. However, the executive order’s vague language on allied collaboration leaves a structural void, effectively locking out the United Kingdom and Australia despite their multibillion‑dollar investments in quantum sensors and autonomous platforms.

The compute shortfall directly hampers AUKUS Pillar II, which relies on training foundation models that ingest sonar, radar, and quantum‑derived data at exaflop scales. Allies currently face three unattractive paths: constructing their own exascale facilities—a multi‑year, multi‑billion endeavor; relying on commercial clouds that demand bespoke security arrangements; or seeking case‑by‑case U.S. access, which lacks a predictable allocation mechanism. Each option fragments research, inflates costs, and slows the rapid iteration cycle essential for maintaining a technological edge over China, which already dominates most critical tech domains.

A pragmatic remedy lies within the existing Genesis framework: extend the CRDA authorities to AUKUS partners under bilateral agreements mirroring current U.S. industry contracts. Such extensions would preserve IP rights, enforce export controls, and embed allied datasets within secure compartments, ensuring sovereign control over trained models. By institutionalizing shared compute, the alliance could leverage U.S. infrastructure while fostering joint innovation, creating a template for future collaborative resources beyond compute—such as data repositories and test ranges. This approach not only safeguards strategic autonomy but also prevents duplicated effort, reinforcing the collective defense posture the AUKUS partnership was designed to achieve.